What causes the "sharp end effect" in the recall of disaster reports?

Master thesis:

What causes the “sharp end effect” in the recall of disaster reports?

Lea Berkemeier s1811800 27/02/2021

University of Twente

First Supervisor: Prof. Dr. J.M.C. Schraagen Second Supervisor: Dr. S. Borsci

Faculty: Behavioral, Management and Social sciences (BMS)

Department: Cognitive Psychology and Ergonomics (CPE)

Abstract

When people are asked to recall how disasters happened, they tend to remember most vividly and frequently the causes that were spatially and temporally close to the disaster itself, called sharp end factors. Until now, why this so-called sharp end effect occurs remained unclear. The current study investigated whether the blaming tendency of a person, the number of sharp ends mentioned, or a person’s locus of control can be used to explain the sharp end effect. Eighty-three participants took part in a study wherein they had to recall three disaster stories, both directly and after one and three weeks, without reading the stories again. After the final recall, participants rated which factors contributed to the incident the most and filled in a locus of control questionnaire. Results indicated that participants in the Condition ‘Blunt end blaming present’ recalled significantly fewer sharp ends and blunt ends than participants in the Condition ‘Blunt end blaming absent’. The sharp end effect was still present regardless of blunt end blaming manipulations. Additionally, a blunt end effect in terms of blaming tendency was found regardless of blunt end blaming or sharp end manipulations. Lastly, participants’ locus of control was found to have no significant influence on recall or blaming tendency. The results of the current study do not give a clear explanation of the sharp end effect, but it was shown that recall and blaming of sharp ends and blunt ends are separate processes. For future research, the addition of sharp end blaming to the disaster stories should be investigated and the responsibility questions should be separately presented per factor.

Keywords: recall, disaster, sharp end, blunt end, sharp end effect, blaming tendency, locus

of control

What causes the “sharp end” effect in the recall of disaster reports?

Disasters often involve complex and long sequences of events that have been examined by accident investigations, accident reconstructions, and accident recall. Research on accident investigation has shown that the grasping of a disaster’s causes proves difficult for investigators and often results in reporting and fixing causes that are most proximate to the disaster (Cedergren & Petersen, 2011;

Lundberg, Rollenhagen & Hollnagel, 2009; Lundberg, Rollenhagen, & Hollnagel, 2010;

Rollenhagen, Westerlund, Lundberg, & Hollnagel, 2010). Furthermore, research by Wurster (2013) and Verschuur (2013) on accident reconstruction has also shown this tendency to report and focus on proximate disaster causes. The two authors investigated both popular and scientific publications on the Chernobyl nuclear power plant disaster (1986) and the Tenerife airport disaster (1977). Additionally, when people are asked to recount how disasters such as the space shuttle Challenger (1986) or the Tenerife airport disaster (1977) came about, they tend to remember most vividly and frequently the causes that were spatially and temporally close to the disaster itself (Moning, 2014). For example, the cold O-rings of the space shuttle or the actions of the KLM pilot, respectively. Little attention has been paid to what could be possible explanations for this effect of focusing on proximate disaster causes. This so-called “sharp end effect” is a very specific phenomenon that has only been demonstrated so far with disaster recall (Moning, 2014).

This report will start by explaining the distinction between the two types of disaster causes, namely blunt end and sharp end causes. Furthermore, previous research on accident investigation, accident reconstruction, and accident recall will be discussed. Several theories of blaming will be presented and used to provide a possible explanation of the sharp end effect. Next, the concept of locus of control will be introduced and included as an additional exploratory variable for the sharp end effect. Lastly, the different paragraphs will be summarized in terms of their relevance for the current research, its research question and hypotheses will be proposed.

Literature review

Sharp ends versus blunt ends

Before going more into depth on possible explanations for the sharp end effect, it is

important to understand the distinction between blunt end and sharp end factors or causes of

disasters. A visualization of this distinction can be found in Figure 1. In general, several factors

can influence an actual disaster. For example, the main factors usually consist of institutional,

organizational, work environment, team, individual, and task factors (Vincent, Taylor-Adams, &

Stanhope, 1998). Each main factor has several contributory factors, for instance, economic pressures, financial priorities, staffing, communication, knowledge, or task design to name some (Rasmussen, 1997). The distinction between sharp end and blunt end factors is quite common across various models, such as the latent failure model of Reason (1990) or the accident causation model of Rasmussen (1997) and Svedung and Rasmussen (2002).

Figure 1

Factors influencing the development of a disaster from Hollnagel (2002)

According to Reason (1997), active failures refer to actions produced by front-line professionals that are directly involved with the process at hand, thus they are a factor at the sharp end. The consequences of actions at the sharp end are often immediately obvious (Besnard &

Hollnagel, 2014). These operators at the sharp end intend to protect the system from their own and others’ errors. Even though operators at the sharp end conduct errors themselves from time to time, they also provide the needed resilience and expertise to imperfect technical systems with their non- technical skills (Flin & O’Connor, 2017; Reason, 1997) According to Reason (1990, p. 173)

“[r]ather than being the main instigators of an accident, [sharp end] operators tend to be inheritors

of system defects created by poor design, incorrect installation, faulty maintenance, and bad

management decisions. Their part is usually that of adding a final garnish to a lethal brew whose ingredients have already been long in the cooking”. For instance, a pilot or surgeon making an erroneous decision resulting in an accident would be considered a sharp end cause. In the current study, sharp end factors are defined as aspects of people that are temporally and spatially close to the disaster.

In contrast, latent failures refer to actions that are produced in earlier stages by operators hidden and further removed from the active end, with consequences of their actions being less obvious. According to Reason (1990, p.173), “the adverse consequences [of latent failures] may lie dormant within the system for a long time, only becoming evident when they combine with other factors to breach the system’s defenses”. For instance, an organization’s safety culture, time pressure, or even government regulations that contributed to the accident would be considered blunt end causes. Yet, blunt end factors are still crucial because they can lead to an outcome failure by constraining sharp end factors (Lundberg et al., 2009; Reason, 1990). In the current study, blunt end factors are defined as aspects of objects, people, or the underlying organization that are temporally and spatially further away from the disaster.

Rasmussen (1997) also points out the complexity and dynamics of adaptive socio-technical systems that ultimately result in disasters. These adaptive socio-technical systems are surrounded by competitive environments or regulatory conditions which are all grouped under blunt end factors. For the individual decision-maker of such a system as well as for readers of disasters, it is often difficult to grasp the full picture or dynamic flow of the events (Rasmussen, 1997). During disasters, usually, a complex sequence of events is happening with both blunt end and sharp end factors being involved to a greater or lesser extent. Several preliminary attempts have been made to explain the sharp end effect, by reviewing and investigating accidents, publications on those accidents, and recall of accidents, all of which will be discussed in the following.

Accident investigation

In accident investigation, investigators try to apply existing accident models by looking at

what led to the accident itself, which factors were playing a role, and what recommendations can

be done to prevent future accidents (Lundberg et al., 2009). The typical approach when

investigating causes of accidents or disasters is the so-called What-You-Look-For-Is-What-You-

Find (WYLFIWYF), which is a process with which causes first get identified and fixed by remedial

actions (Hollnagel, 2008; Lundberg et al., 2009). However, due to the increasing complexity of

disasters, this usual WYLFIWYF principle is not always applicable. It often results in trying to fix specific, individual problems close to the disaster that was found during the investigation (Lundberg et al., 2009).

Similar results were also found by Cedergren and Petersen (2011), showing that accident investigations considerably emphasize causes at the micro-level such as physical processes, equipment, or actor activities. In turn, investigators pay less attention to the meso-level, such as organizational factors or macro-level such as regulations of the government. Cedergren and Petersen (2011) concluded that greater diversity among investigators in terms of technical and operational background is needed to enhance the scope on what factors are causes and how the different factors interacted. Even though this change of perspective did take place in accident investigations, an emphasis on human error is still visible (Dekker, Nyce & Myers, 2013). Research by Rollenhagen et al. (2010) found that technical factors were often not seen as a salient cause, but that the samples rather perceived the technologies’ performance as expected and considered technical weaknesses as a symptom of non-technical factors. In sum, although several attempts have been made at reducing the emphasis or focus on sharp end factors during accident investigations, this focus on sharp ends is still very prominent. How this tendency is reflected in the actual publication of the disasters will be discussed in the following part.

Accident reconstruction

In general, people tend to deal with complexity through oversimplification (Feltovich, Hoffman, Woods & Roesler, 2004). Research by Feltovich et al. (2004) has shown that people tend to reduce complex information into simple and understandable components, the so-called reductive tendency when being asked to reproduce information. Geurts (2013) explained that authors are intentionally reducing disaster descriptions because they do not regard every information as necessary to mention as it seems either too obvious to mention or is not necessary to support their argument. Furthermore, news media often make use of framing, meaning that the media tends to select things such as visual images or journalistic analysis to highlight some aspects of an event while ignoring or downplaying others (Druckman 2001; Haider-Markel & Joslyn 2001; Iyengar 1991; Scheufele, 1999). As a result, the disaster descriptions that one reads in a report or the news are already a reduction of the complex reality of the disaster. Thus, the reductive tendency of disaster reports could influence what people remember as they do not get presented with all facts.

Of course, it is frequently not feasible to do so due to, for instance, word limits imposed on news

media. However, it is important to create a consciousness among people that any description or report of a disaster is also a simplification of what happened and vary among authors, to prevent future distortions (see Geurts, 2013; Vicente & Brewer, 1993).

Verschuur (2013) conducted a literature review on 31 non-scientific and 31 scientific articles reports of the Chernobyl disaster with a minimum length of 100 words and a maximum length of 500 words. Similarly, Wurster (2013) conducted a literature review on 38 scientific and 29 non-scientific disaster reports of the Tenerife disaster with a minimum length of 100 words and a maximum length of 500 words. Both Verschuur (2013) and Wurster (2013) found a high information reduction regarding the causes of the accident over time. Additionally, the focus on sharp end factors was significantly higher for non-scientific articles than for scientific articles of Chernobyl (Verschuur, 2013). Verschuur (2013) suggested that this focus is either due to the story- like construction of non-scientific articles which makes them interesting to read or it is due to the human factor at the sharp end. The concept of human error was mentioned more often in non- scientific articles which might be due to the Western culture, more specifically due to the personal responsibility in failure and achievement (Dekker & Nyce, 2012; Verschuur, 2013). In contrast, Wurster (2013) found that sharp end factors were mentioned significantly more often in both scientific and non-scientific disaster reports of Tenerife than blunt end factors.

Geurts (2013) conducted a literature review on 21 scientific and 17 non-scientific disaster reports of the Challenger disaster with a minimum length of 50 words and a maximum length of 500 words. In contrast to the results of Verschuur (2013) and Wurster (2013), Geurts (2013) did not find a significant difference in the occurrence of sharp end and blunt end factors in publications of the Challenger disaster. One reason for these results could be that there were no sharp end factors involved in the decision making or in the Challenger disaster itself. The decision making was made the day before by other people than the crew and the sharp end factors temporally close to the disaster were more technically-related such as the O-rings that froze due to low temperatures. In contrast, in the Tenerife disaster, the cockpit crew and in the Chernobyl disaster the nuclear power plant operators were both spatially and temporally close to the disaster and involved in the decision making.

As a reduction of information is impossible to avoid, Geurts (2013) suggested that one

should be careful with such reports to avoid misconceptions. For instance, one can reduce this by

using original investigation reports instead of secondhand sources (with the caveat mentioned in

the previous section that accident reports also suffer from a focus on sharp ends). Building upon the distinction made by Vicente and Brewer (1993) between the accuracy and completeness of the original sources and the subsequent recall of those sources, a likely explanation for the focus on sharp ends might be due to recall processes. That is, even when sources present a well-balanced account of both blunt ends and sharp ends, selective encoding might lead to an unbalanced recall of sharp ends. In the following, another line of research will be discussed on what people are recalling of disaster stories after reading them.

Accident recall

Moning (2014) investigated the effects of story grammar, thus the predetermined order or leitmotiv of a text, on people’s recall of sharp end and blunt end factors after reading a disaster report. Participants read two disaster stories with a length of around 300 words, one on the Challenger and one on the Tenerife disaster (Moning, 2014). Both disasters varied according to their story grammar, thus whether a story grammar was present in the report or not. Furthermore, the stories varied in the number of sharp ends and blunt ends mentioned, including either four sharp end causes and two blunt end causes or vice versa. The combination of the two variables, story grammar and the number of sharp end and blunt end causes, led in total to four different versions of the Tenerife and four of the Challenger disaster. After reading each story, participants filled out a crossword puzzle intended to erase information in working memory and were then asked to write everything down that they remembered of the article they just read (Moning, 2014).

The results showed, first and foremost, a sharp end effect, namely that participants recalled significantly more sharp end than blunt end causes, regardless of the number of sharp ends and blunt ends included in the texts. Additionally, the number of recalled sharp end and blunt end causes was significantly lower for the Challenger disaster than for the Tenerife disaster (Moning, 2014). Moning (2014) suggested that these differences were due to the Challenger disaster being more technical and therefore, more difficult to comprehend than the Tenerife disaster. This technical complexity might also serve as another explanation of the results found by Geurts (2013).

Besides, this sharp end effect occurred regardless of whether a story grammar was present or not

(Moning, 2014). Therefore, the mere presence of a story grammar cannot be used to explain the

sharp end effect for recalling disaster-related information or causes. Furthermore, a main effect of

story grammar was found by Moning (2014), such that participants remembered and recalled more

information when a story grammar was present. This effect of story grammar on memory was

previously reported by Mandler and Johnson (1977) and Thorndyke (1977), with stories that conformed more to an ideal story grammar structure being better recalled than those deviating from it.

Several limitations of Moning’s study (2014) are worth mentioning, namely, that recall was conducted only at one point in time and that it has not been tested what happens when the disaster report contains an equal number of sharp ends and blunt ends. Furthermore, the Tenerife and Challenger disaster are both complex disasters for readers that are not familiar with disasters. The current study aims to control for these limitations by choosing for disasters with fewer technical or complex details. Additionally, the recall of different disasters will be measured over several weeks, to establish the robustness of the sharp end effect.

The emphasis on the sharp end factors can be explained by the Western moral enterprise but also because finding a cause of or a factor that is blameworthy for an accident is inherent to human nature (Dekker & Nyce, 2012; Monroe & Malle, 2019). In turn, the blaming tendency of a person likely influences what she or he devotes attention to and thus results in different elements of disasters getting stored and recalled. Thus, it could be that the sharp end effect in terms of recall, as shown by the study of Moning (2014), is because people blame those most closely in time and space to the actual outcome of the accident. In the following paragraph, this inherent blaming tendency of humans will be explained and an explicit connection made with the memory phenomenon of the sharp end effect.

Blaming tendency

Research within social psychology has suggested that people make initial, quick attributions based on their prior beliefs or experiences (Anderson, Krull, & Weiner, 1996). When people are presented with multiple explanations, they tend to accept the explanation that is most plausible to them and reject those seeming implausible based on their accessible knowledge structures such as intuitive causal theories (Anderson et al., 1996). As stated by Rasmussen (1997), accidents are often attributed to human errors and independent failures, which is often an inadequate conclusion regarding the actual causes of the accident. It has been suggested that people tend to blame more the people that were closest to producing but also to possibly avoiding the accident as it is more emotionally satisfying and convenient (Beso, Franklin, & Barber, 2005).

Furthermore, Hollnagel (2004) also explained that with a safety culture focusing on mistakes and

sanctions, blaming someone creates a certain authority or power which makes management prone

to blame the bottom of the hierarchy, namely the sharp end. Therefore, one possible explanation of the sharp end effect could be that people recall sharp ends better because blaming a specific human operator or person seems more plausible and easier to do than blaming a complex system or organization.

Furthermore, it has been proposed within the area of philosophy that people tend to scapegoat other people or a group of people due to reasons such as maintaining a moral value or perceived personal control (Rothschild, Landau, Sullivan & Keefer, 2012). Scapegoating is defined as “the act of blaming and often punishing a person or a group for a negative outcome that is due, at least in large part, to other causes” (Rothschild et al., 2012, p. 1148). With scapegoating, people try to minimize feelings of guilt regarding their own responsibility for a negative outcome by transferring the blame towards other individuals or a group (Douglas, 1995). Additionally, it has been shown that people tend to “externalize blame for negative outcomes that would otherwise incriminate themselves or their group” (Rothschild et al., 2012, p. 1149).

Scapegoating serves as a strategy to maintain the image of an orderly, stable, and predictable external world (Allport, 1948). According to Lagnado and Channon (2008), people are seeking causal explanations for how and why things happen in their daily life. Scapegoating helps people to restore perceived control and provides them with a simple explanation of the event as scapegoats can be clearly identified in contrast to the usual chaotic factors (Rothschild et al., 2012).

The success of scapegoating depends on whether a person perceives the target as a viable or nonviable scapegoat based on the target’s ability and intent to deliberately cause the outcome or not (Glick, 2005). Even if there is no plausible reason, “it is possible that people prefer seeing viable scapegoats as responsible for a seemingly random negative outcome to leaving that outcome unexplained” (Rothschild et al., 2012, p. 1149). Based on these suggestions, the current study will investigate whether a person’s blaming tendency can be influenced or changed when offering the reader a scapegoat to blame, in this case, the blunt end factors.

According to the culpable control model of Alicke (2000), it is assumed that people’s

spontaneous and quick assessment or evaluations encourage blame judgments. According to

Lagnado and Channon (2008, p. 757), “[t]hese evaluations are less deliberative than judgments of

personal control and can lead to significant biases in the processing of relevant information. In

particular, they typically result in greater blame being ascribed to human agents, and less notice

taken of mitigating circumstances”. Furthermore, Alicke, Davis, and Pezzo (1994) found that this

greater ascription of the blame also resulted in distorted causal control assessments that people used to justify their blame attributions. In many cases, the people rather than environmental factors are perceived as the primary controlling forces that underlie negative evidence (Cook & Woods, 1994; Jones, 1990). This perception is based on the temporal and physical relationship of the human operators to the outcome (Cook & Woods, 1994). Human actions usually seem more controllable, easier to imagine and are often the abnormal feature in ordinary situations (Alicke, 2000;

Kahneman & Miller, 1986; Hart & Honoré, 1985). However, it remains unclear whether causal attributions are made to factors that are at the beginning of the causal chain, thus a primacy effect, or to factors that are directed to the factors at the end of the causal chain, thus a recency effect (Einhorn & Hogarth, 1986; Miller & Gunasegaram, 1990; Johnson, Ogawa, Delforge, & Early, 1989; N’gbala & Branscombe, 1995; Vinokur & Ajzen, 1982). However, according to Alicke (2000), the more proximate the action and its effect are, the greater the causal control over the outcome is and results in a higher degree of causality. Therefore, the current study investigates whether removing sharp end factors from the disaster stories, thus the factors that are suggested to receive the blame, will influence both recall and blaming tendency.

In philosophy, two perspectives on blaming are proposed which differ in the way that

people attend new incoming information and how or whether their blame judgment changes,

namely the socially regulated blame perspective and the motivated-blame model. It is worth

mentioning that both theories are important and compatible as they apply to different conditions of

blaming (Monroe & Malle, 2019). According to the socially regulated blame perspective, blaming

is usually a socially regulated process with people systematically attending and processing blame-

relevant information (Monroe & Malle, 2019). Offering a warrant is a social demand required when

people blame others, as it provides them with evidence that one’s moral judgments are justified

and fair (Monroe & Malle, 2019; Voiklis & Malle, 2018). As the research of Malle (2021) has

shown, the best way to support warrants for blame judgments can be done by presenting

information that is usually already processed by the reader to form a blame judgment, such as

causality, reasons, and evidence supporting these inferences. If the reader gets new information, it

is predicted that they flexibly revise their blame judgments if the information is meaningful to them

(Monroe & Malle, 2019). Thus, by publicly expressing blame, these blame judgments are expected

to become more nuanced and systematic (Monroe & Malle, 2019).

According to the motivated-blame model, blaming is considered as an inherent desire that is driven by people’s intuitive emotional responses and their need to rationalize and explain norm- violating behaviors (Greene, 2008; Monroe & Malle, 2019). Therefore, their need to find someone to blame is biased in that they favor information that confirms their existing blame judgment over information that would mitigate the blame (Alicke, 2000; Ames & Fiske, 2013). It has been suggested that negative evaluations or spontaneous reactions lead people to see the source as blameworthy and interpret available evidence in a way to support this blaming hypothesis (Alicke, Rose, & Bloom, 2011).

In summary, the main difference between the two perspectives is to what extent the person is willing to adjust her or his initial blame judgment based on new incoming and relevant information. In the study, the focus will be on the socially regulated perspective, because it will be tested whether the readers quickly adjust their initial blame judgment when presenting them with blunt end blaming. As stated by Monroe & Malle (2019), the blame judgment of a person influences what kind of information she or he pays attention to. In turn, this will also influence what kind of information is getting processed, stored, and consequently recalled. For the current study, it will be investigated whether the underlying processes of blaming tendency and recall are connected or separated. If the blaming tendency and recall underlie connected processes, then the addition of blunt end blaming will result in both higher blame towards blunt end factors and a higher blunt end recall. In contrast, if the blaming tendency and recall underlie separate processes, then the addition of blunt end blaming will result, for instance, in higher blame towards blunt end factors but will not affect or lower blunt end recall. With our initial assumption that people recall sharp ends better because they assign more blame to sharp end factors, it will be expected that the blaming tendency and recall underlie connected processes. Thus, by the addition of blunt end blaming it is expected that people will revise their initial sharp end blame judgment to blunt end blame judgment. In turn, this will affect their information processing and ultimately result in blunt end factors getting better recalled than sharp end factors.

Whether scapegoating is successful and restores perceived control likely also depends on

the individual predisposition or personality traits. For instance, in how far or to what extent the

person thinks the environment can be influenced or controlled, in other words, the locus of control

of a person (Rotter, 1966). In the following, the concept of locus of control will be explained and

discussed.

Locus of control

According to Wang and Lv (2017, p. 2339), locus of control is defined as “a generalized enduring expectancy or belief about how responsive and controllable the environment is”. Overall, a distinction is made between internal and external locus of control, however, it can also be the case that the tendency is equal between these two. The type of locus of control a person has depends on whether the person sees a causal relationship between her or his own behavior and the reward or not (Rotter, 1966). When people have an internal locus of control, they usually “believe that the environment is responsive to their own relatively permanent characteristics and that rewards are contingent on personal actions […]” (Wang & Lv, 2017). Additionally, they tend to see themselves as responsible for the outcomes of their actions (Suárez-Álvarez, Pedrosa, García-Cueto, & Muñiz, 2016). In turn, people with an external locus of control tend to see external rewards and their environment as uncontrollable. These people tend to give credit to external factors such as coincidence or luck if their actions turn out to be successful or unsuccessful (Rotter, 1966; Weiner, 1979). Taking this further, it could be that the type of locus of control, thus the view one has on the events or environment, influences or directs the reader's attention to particular aspects of stories.

In other words, it could be that people with an external locus of control, or an equal tendency focus on and recall different aspects of a story, for instance, the number of sharp ends or blunt ends, than people with an internal locus of control.

An important distinction should be made here, namely between attributional explanations which are mostly post hoc, while locus of control is more about the prediction of the ability one has to control the future (Galvin, Randel, Collins & Johnson, 2018; Ng, Sorensen, & Eby, 2006).

More specifically, attribution theory focuses on the causal inferences people make (Heider, 1958;

Kelley, 1973; Weiner, 1986). Thus, attributions and locus of control can be aligned or misaligned.

For instance, a person may think that the illness she or he experiences is due to factors beyond her or his own control (external attribution) but believes that they can exert some control to promote recovery (internal locus of control; see White, Lehman, Hemphill, Mandel, & Lehman, 2006).

Based on these findings and the scapegoat theory, it is hypothesized that people with an external

locus of control or an equal tendency have a different blaming tendency as they attribute the cause

to a different scapegoat than people with an internal locus of control.

Current study

The main purpose of the current study is to investigate possible causes of the sharp end effect in the recall of disaster reports. The study intends to replicate the work of Moning (2014) as the main framework, but at the same time also adds certain changes and extensions to it.

Additionally, it will be investigated whether or not there is a connection between the recall and blaming tendency of sharp ends and blunt ends. With the extensions of the current study, it will be explored whether people tend to recall sharp ends more frequently because they search for someone, thus a person, to blame rather than something or an organization to blame. To answer the research question of “What causes the sharp end effect in the recall of disaster reports?”, the following hypotheses will be tested.

Recall

The current study will explore how people recall which percentages of sharp ends and blunt ends with different types of conditions. Based on the socially regulated blame perspective of Monroe and Malle (2019) and the preliminary findings of Moning (2014), it is predicted that the addition of blunt end blaming will affect information processing and result in a different recall of texts. If blunt end blaming is added to the texts, it is expected that the reader’s attention will be shifted towards blunt ends and result in a poorer recall of sharp ends due to systematically attending and processing blame-relevant information. The addition of blunt end blaming is also expected to result in an increased blunt end recall in comparison to the absence of blunt end blaming due to the same argumentation. Based on the research of Lagnado and Channon (2008), it is predicted that the sharp end removal will reverse people’s quick evaluations of situations and thus influence both processing and recall of information. Through the sharp end removal, it is expected that people will pay more attention to the mitigating circumstances, the blunt ends, which will result in an increased recall of blunt ends than for texts with sharp ends present.

Most studies, including the research of Moning (2014), just measure recall at one point in

time. There is no literature available on the recall of sharp ends and blunt ends over weeks with the

current manipulations of blunt end blaming and number of sharp ends. Therefore, no directed

hypotheses for the effects over weeks can be posited. Going further, as disasters usually always

involve sharp ends and blunt ends, the sharp end removal will likely result in a different recall order

than for disasters where sharp ends are mentioned. Also, it is predicted that the addition of blunt

end blaming will result in a different recall order than the absence of blunt end blaming. As there

is no literature available on the differential recall order of sharp ends and blunt ends, we cannot posit a directed hypothesis for the effect on recall order. The following set of hypotheses was posed to investigate the recall of sharp ends and blunt ends:

1. The recall of sharp ends will be higher for participants who read texts without blunt end blaming at the end than for participants who read texts with blunt end blaming at the end 2. The sharp end recall will differ over time between participants who read texts with blunt

end blaming at the end and participants who read texts without blunt end blaming at the end 3. The recall order of sharp ends will differ between participants who read texts with blunt end blaming at the end and participants who read texts without blunt end blaming at the end 4. The recall of blunt ends will be higher for participants who read texts without sharp ends

than for participants who read texts with sharp ends

5. The recall of blunt ends will be higher for participants who read texts with blunt end blaming at the end than for participants who read texts without blunt end blaming at the end 6. The blunt end recall will differ over time between participants who read texts with sharp

ends and participants who read texts without sharp ends

7. The blunt end recall will differ over time between participants who read texts with blunt end blaming at the end and participants who read texts without blunt end blaming at the end 8. The recall order of blunt ends will differ between participants who read texts with sharp

ends and participants who read texts without sharp ends

9. The recall order of blunt ends will differ between participants who read texts with blunt end blaming at the end and participants who read texts without blunt end blaming at the end

Blaming tendency

Based on the assumptions of the scapegoat theory (Rothschild et al., 2012; Glick, 2005) and the socially regulated blame perspective (Malle, 2021; Monroe & Malle, 2019; Voiklis &

Malle, 2018), it is predicted that the addition of blunt end blaming and removal of sharp ends will

influence reader’s blaming tendencies towards sharp ends and blunt ends. The addition of blunt

end blaming will likely increase the blame towards blunt ends, as the readers are offered a warrant

and viable scapegoat in comparison to texts without blunt end blaming. Due to the same

argumentation, the addition of blunt end blaming will likely decrease the blame towards sharp ends

in comparison to texts without blunt end blaming. The sharp end removal will likely result in an

increased blunt end blame as the blunt end is the only available viable scapegoat mentioned in the disaster. In turn, the sharp end removal will likely also result in a decreased sharp end blame due to the same argumentation. The following set of hypotheses was posed to investigate the blaming tendency towards sharp ends and blunt ends:

10. The assigned blame towards sharp end factors will be higher for participants who read texts with sharp ends than for participants who read texts without sharp ends

11. The assigned blame towards sharp end factors will be higher for participants who read texts without blunt end blaming at the end than for participants who read texts with blunt end blaming at the end

12. The assigned blame towards blunt end factors will be higher for participants who read texts without sharp ends than for participants who read texts with sharp ends

13. The assigned blame towards blunt end factors will be higher for participants who read texts with blunt end blaming at the end than for participants who read texts without blunt end blaming at the end

Locus of control

Lastly, it will be explored whether a person’s locus of control directs one’s attention more to blunt end or sharp end factors and thus result in an increased or decreased recall (Rotter, 1966).

It will also be examined whether a person’s locus of control influences the blaming tendency, as they could differ in whether they perceive the sharp end or blunt end factors as viable scapegoats (Glick, 2005; White, Lehman, Hemphill, Mandel, & Lehman, 2006). No directed hypotheses will be posited regarding locus of control, given the absence of relevant literature. The following set of hypotheses was posed to investigate the blaming tendency towards sharp ends and blunt ends:

14. There will be a difference in sharp end recall based on the participant’s locus of control 15. There will be a difference in the assigned blame towards sharp end factors based on the

participant’s locus of control

16. There will be a difference in blunt end recall based on the participant’s locus of control 17. There will be a difference in the assigned blame towards blunt end factors based on the

participant’s locus of control

Method Participants

Ninety-six students currently or previously enrolled at universities participated in the experiment. For the experiment, a sufficient understanding of English and the ability to answer questions written in English was required, as participants had to read, recall and answer questions about disaster stories written in English. This was ensured by only including participants who completed their A-level with English courses taken until the final A-level year. In other words, participants’ English level had to be between B2 and C1, which universities with English study programs usually require for enrolment. If participants were not enrolled in an English study program, they were judged on a question about how they perceive their English reading and understanding skills (see Appendix A). Participants that had foreknowledge about the disasters were excluded from the study. Three participants turned out to be not suitable for the experiment, as they had in-depth knowledge about at least one of the disasters. Another five participants were excluded from the results as well as they dropped out after the first or second session of the experiment. Furthermore, three participants were excluded because they indicated on a question that was asked at the very end of the experiment that they searched for information on the Internet about the disasters during the experiment (see Appendix A). Lastly, two participants were left out due to extremely low recall scores compared to the rest of the sample and were identified as outliers, leaving 83 valid cases for the current study.

Description of the sample

Regarding gender, 27 (32.5%) were male, 55 (66.3%) were female and one participant identified as another gender (1.2%). The age ranged between 21 and 35 years, with a mean age of 24.22 years (SD = 2.38). Furthermore, 27 (32.5%) participants had a Dutch nationality, 40 (48.2%) had German nationality and 16 (19.3%) had another nationality. In total, 72 (86.7%) participants indicated that they are or were enrolled in a study program taught in English at a university or university of applied sciences, 8 (9.6%) were temporarily enrolled and 3 (3.6%) were not enrolled.

Participants that were not or only temporarily enrolled rated their ability to read English texts or

stories as sufficient (rating at least 4 out of 7, see Appendix A). No differences were found in terms

of content and length of the written texts in comparison to participants enrolled in a study program

taught in English. Therefore, these participants were included in the study. Participants were

recruited through convenience sampling. Before the experiment, the participants needed to give

their informed consent for participation (see Appendix B), in accordance with ethical guidelines and procedures of the University of Twente. The study was approved by the BMS Ethical Committee of the University of Twente.

Materials

As participants had to read and recall stories about different disasters, several research articles were searched and retrieved from the databases of www.google.com and www.google- scholar.com. The requirements for the search were that both scientific and non-scientific publications were available for the disaster and the disaster needed to consist of at least two blunt end and two sharp end factors. Three accidents were chosen, namely the train disaster of Eschede (1998) in Germany, the Kiss Nightclub Disaster (2013) in Brazil, and the WHO concert disaster (1979) in the USA. Both scientific and non-scientific publications can be found in Appendix C.

Participants received three different shortened disaster reports that were based on the original

disasters of Eschede, Kiss Nightclub, and WHO concert with a length of a maximum of 350 words

(see Appendix D). The original texts were partly changed for the purpose of the study. For instance,

fictional causes were added to the stories, to arrive at an equal number of four sharp ends and four

blunt ends. Each disaster consisted of four different versions, differing in whether blunt end

blaming was present or absent and in the number of sharp end causes mentioned either four or none

(see Appendix D). The number of words or sentences missing when no blunt end blaming or sharp

ends were mentioned in the text was filled up with additional neutral, contextual information that

was neither related to blunt- nor sharp ends. The different versions of each disaster were controlled

for readability and the level of abstractness and concreteness (see Appendix D). For readability,

ensuring that they have the same length of sentences and number of words, the text was run through

an algorithm (see http://www.readabilityofwikipedia.com). To compare readability across the

different versions of the disasters, the Flesch-scores were calculated to indicate how difficult it is

for a reader to understand the selected English passages. The Flesch-scores of all articles were

between 57 and 73, indicating that the different disaster stories had a good and similar reading level

(see Appendix D). For determining and controlling the level of abstractness and concreteness, the

number of concrete and abstract words was counted with help of a validated list of concrete and

abstract words (Brysbaert, Warriner & Kuperman, 2014), and the percentages for each disaster

story calculated (see Appendix D). The deviations in terms of level of abstractness and

concreteness across the different disaster versions were kept at a maximum of 5%.

The Internal-External Locus of Control Scale of Rotter (1966) was used to determine the participant’s locus of control at the end of the experiment. According to Wang and Lv (2017), the scale is most widely used and has food psychometric properties, with good average reliability of 0.70 (Ng et al., 2006). Furthermore, the scale correlates well with other methods to assess the locus of control (Wang & Lv, 2017). It consists of 29 forced-choice items, including six filler times to make the purpose of the test more ambiguous (see Appendix E).

Coding schemes

The coding scheme of Moning (2014) was used as an example and then adjusted to the criteria of the current study. For each condition, a slightly different coding scheme was developed as the manipulations resulted in a slightly different categorization of information and text order (see Appendix F). For the Condition ‘Blunt end blaming present/Sharp ends present’, four categories were created, namely contextual information, sharp end factors, blunt end factors, and charges towards the blunt end mentioned. For the Condition ‘Blunt end blaming absent/Sharp ends present’, three categories were created, namely contextual information, sharp end factors, and blunt end factors mentioned. For the Condition ‘Blunt end blaming present/Sharp ends absent’, three categories were created, namely contextual information, blunt end factors, and charges towards the blunt end mentioned. For the Condition ‘Blunt end blaming absent/Sharp ends absent’, two categories were created, namely contextual information and blunt end factors mentioned.

Four different second-raters were assigned to different conditions, where each filled in the coding schemes of eight participants. Cohen’s κ was run to determine if there was agreement on whether the coding schemes for the different conditions would be filled in similarly. According to the guidelines of Landis and Koch (1977), there was almost perfect agreement in terms of Condition ‘Blunt end blaming present/Sharp ends present’ coding schemes, κ = .868 (p < .001), substantial agreement in terms of Condition ‘Blunt end blaming absent/Sharp ends present’ coding schemes κ = .708 (p < .001), Condition ‘Blunt end blaming present/Sharp ends absent’ coding schemes κ = .787 (p < .001) and Condition ‘Blunt end blaming absent/Sharp ends absent’ coding schemes κ = .797 (p < .001).

Design

A between-subjects design with three measurement points in time was chosen since the

recall and assigned blame scores of participants in the four conditions were compared with each

other. The four conditions resulted from a two-by-two fully factorial design, which can be found in Table 1. The independent variables were the two manipulations namely Sharp ends (present/absent) and Blunt end blaming (present/absent). Note that we chose for the selective removal of sharp ends rather than blunt ends, as we already had preliminary evidence for the existence of a sharp end effect (Moning, 2014), and would obtain more conclusive evidence if we were able to eliminate the sharp end effect. We manipulated the inclusion of blunt end blaming rather than sharp end blaming for the same reasons as mentioned above with the manipulation of sharp ends in the text.

The recall and recall order were measured with a free recall task over three points in time.

In the study of Moning (2014), a free recall task was used by asking participants to write down as many aspects as they can recall from reading about the disaster. Similar studies about recall by Nevid, Pyun and Cheney (2016) and Rohrer and Pashler (2010) also made use of a free recall task.

To test the effects of retention intervals of disaster stories, the recall was measured with three data points to see whether the sharp end effect stays the same over the weeks. Sharp end and blunt end causes were determined based on the categorization of different bachelor students (Moning, 2014;

Wurster, 2013; Geurts, 2013). The dependent variables of Sharp end recall and Blunt end recall

were averaged percentages, calculated by dividing the number of recalled sharp and blunt end

elements by the number of sharp and blunt ends that could have been recalled. The dependent

variables of Sharp end recall order and Blunt end recall order were the averaged delta differences

in the number of sentences between the original order of sharp ends and blunt ends in the text and

the order of participants’ written texts. Lastly, the dependent variables of Assigned blunt end blame

and Assigned sharp end blame were calculated by averaging the assigned blame by participants on

several questions with a scale from 1, being not responsible at all, to 5, being extremely responsible,

of each blunt end and sharp end factor. The to-be-rated sharp end and blunt end factors were all

presented simultaneously per disaster in the responsibility questions. Additionally, a “Not

applicable” option was added to the right side of the scale as an implicit memory test for

participants who read texts without sharp ends (see Appendix G). The first session required on

average 43 minutes (SD = 9.62), the second session around 18 minutes (SD = 7.3), and the last

session around 42 minutes (SD = 50.92).

Table 1

Experimental conditions

Sharp ends

present absent

Blunt end blaming

present Blunt end blaming present/Sharp ends present

Blunt end blaming present/Sharp ends absent absent Blunt end blaming

absent/Sharp ends present

Blunt end blaming absent/Sharp ends absent Note. Blunt end blaming involved the absence or presence of one or two sentences of charges, named towards the blunt end at the end, and blaming words used to describe the actions of the blunt ends in a negative way. Sharp ends involved the absence or presence of four sharp end factors.

Procedure

The experiment was held online via one-on-one Google Meets conversations where participants were asked to share their microphone, video, and screen during the whole session. The structure of the experiment was based on the study of Moning (2014) and implemented with slight additions. A pilot study was conducted before the actual experiment with four participants, one in each condition, to test for understandability and time limitations.

The questions and materials that participants read and filled in were created with Qualtrics.

Participants first filled in an informed consent and provided their email address, which would be later removed from the dataset. Randomization of the story order as well as of the assigned conditions was created in Qualtrics to avoid any confounding effects. The email address was used to reinvite the participants for the follow-up sessions and take previous data such as condition or order into account for the next session. Next, participants answered two questions per disaster story regarding possible foreknowledge of the disasters in written form (see Appendix A) to avoid any confounding effect due to knowledge.

Participants were not informed about the purpose of the total duration of the experiment to avoid any confounding memory effects as some would then prepare for the upcoming sessions.

After randomizing the order of the stories, participants were asked to read the first story twice for

which they had a maximum of five minutes, see Table 2. The reading time of five minutes per story

was based on the median of the four participants of the pilot study. A visible timer on Qualtrics was set, so that the program would automatically jump to the calculation task after five minutes or participants could advance earlier themselves if they were done before five minutes. During the calculation task, participants had to solve several subtraction calculation tasks for two minutes where they were asked to think aloud the whole time. The calculation time was based on the pilot study, where one minute was indicated as too short and three minutes as too long, therefore resulting in two minutes. With the calculation task, it was intended to distract the participants for some time to find out what participants recall without relying on their working memory. Afterward, participants were asked to write down everything that they remembered of the first disaster they read. Next, the same steps were repeated twice with two other disasters. Participants were asked at the end of the first session whether they could come back after one week to fill in a part that had not been developed yet.

After one week, the second session of the experiment took place and participants had to

write down one more time sequentially what they remembered of the stories they read one week

ago without reading them again. The same procedure was repeated after another two weeks during

the third session. Next, participants rated each sharp end and blunt end factor in terms of

contributing to the accidents. Afterward, participants filled in the 29 statements of the Internal-

External Locus of Control Scale (Rotter, 1966). Finally, participants filled in demographical and

background questions, see Appendix G. They also filled in questions where they should indicate

whether their answers to the foreknowledge question were still valid and whether they acquired

any additional information about the disasters while the experiment was going on. At the end of

the last session, participants were debriefed about what the experiment was about.

Table 2

Experimental structure per week

Week First disaster → Second disaster → Third disaster Week 0 1. Reading text

2. Calculation task 3. Recalling text

1. Reading text 2. Calculation task

3. Recalling text

1. Reading text 2. Calculation task

3. Recalling text Week 1 1. Recalling text 1. Recalling text 1. Recalling text Week 3 1. Recalling text 1. Recalling text 1. Recalling text 2. Responsibility

questions 3. Locus of control

questions

4. Demographic questions

Data Analysis

The data were analyzed with IBM SPSS Statistics 26 and a significance level of 0.05 chosen. For the hypotheses regarding sharp end recall and sharp end recall order, a two-way repeated-measures MANOVA was conducted. To investigate blunt end recall and blunt end recall order, a three-way repeated-measures MANOVA was conducted. For the hypotheses regarding blaming tendencies towards sharp- and blunt end factors, a two-way MANOVA was conducted.

Finally, for the hypotheses regarding the locus of control, a one-way MANOVA was conducted.

First, the overall recall of the different disaster stories was investigated and compared with

each other. Then, normality checks were conducted, and the data prepared. Descriptive statistics

for the independent variables (Blunt end blaming, Sharp ends) and dependent variables (Blunt end

recall, Blunt end recall order, Assigned blunt end blame, Sharp end recall, Sharp end recall order,

Assigned sharp end blame) were calculated. Finally, several statistical analyses for testing the

different hypotheses were conducted comparing the effect of sharp end presence or absence and

blunt end blaming presence or absence.

Results Normality checks and data preparation

First, the different dependent variables, namely the number and order of recalled elements as well as assigned blame to sharp end and blunt end factors, were inspected to see whether they deviate from a normal distribution. This was done with help of QQ and stem and leaf plots, which revealed that the data approached a normal distribution with minor skewness. Identified outliers were evaluated on two criteria, namely whether they deviated more than four standard deviations from the mean and whether their absence or presence would induce a large change in the results.

For most outliers that deviated more than four standard deviation, the presence or absence did not induce a large or significant change. However, one participant who read texts with blunt end blaming at the end and without sharp ends was removed due to extremely low values that were listed as outliers in several dependent variables.

Recall of the different disaster stories

The averaged recall scores of all possible elements across the three disasters were very similar, with participants recalling on average 52.7% (SD = 15.91) of the ICE Disaster elements, 54.8% (SD = 12.78) of the Kiss Nightclub Disaster elements, and 51.1% (SD = 14.58) of the ‘The Who’ Concert Disaster elements. Based on these results, no distinctions were made between the disasters in the following parts as they were treated as replications of each other.

Hypotheses recall (1-9) Sharp end recall

Regarding sharp end recall, a significant main effect of time was found [F(1.55, 63.50) = 32.10, p = .000], with a large effect size (partial η

= .44), with the recall dropping from immediate recall (M = .78, SD = .15) to one week delayed recall (M = .65, SD = .18) and stabilizing with three weeks delayed recall (M = .63, SD =.20).

The first hypothesis was that the recall of sharp ends will be higher for participants who

read texts without blunt end blaming at the end than for participants who read texts with blunt end

blaming at the end. With the factorial ANOVA test of between-subjects effects, a statistically

significant difference of sharp end recall in terms of blunt end blaming was found [F(1, 41) = 4.66,

p=.037] with a medium to large effect size (partial η

= .10). In other words, the recall of sharp

ends was significantly higher for participants who read texts without blunt end blaming at the end

(M = .74, SD = .15) than for participants who read texts with blunt end blaming at the end (M = .63, SD = .16). Therefore, the first hypothesis is accepted.

Sharp end recall over the weeks

With the second hypothesis, it was explored whether sharp end recall changes over time with blunt end blaming presence or absence. According to the repeated measures ANOVA with a Greenhouse-Geisser correction, no statistically significant interaction effect was found, see Table 3. Thus, the effect of blunt end blaming presence or absence on sharp end recall stays the same over the three weeks. Therefore, the second hypothesis is rejected.

Table 3

Means and standard deviations of sharp end recall per manipulation from the factorial repeated measures ANOVA of between-subject effects

Sharp end recall Immediate One week

delayed

Three weeks delayed

M SD M SD M SD F p Partial

η

Blunt

end blaming

Present .73 .14 .60 .18 .58 .20 0.04 .933 .00 Absent .82 .13 .70 .16 .69 .20

Sharp end order recall

On average, the disaster stories had a total length of 28 sentences. For the sharp end and

blunt end order recall, the values could range between 0-27 sentences, with a large positive value

indicating that the causes were recalled much later in the participant’s text than the original order

and vice versa. The third hypothesis was that the recall order of sharp ends will differ between

participants who read texts with blunt end blaming at the end and participants who read texts

without blunt end blaming at the end. With the factorial ANOVA test of between-subject effect,

no statistically significant difference of sharp end recall order in terms of blunt end blaming was

found [F(1, 41) = 0.56, p = .461] with a small effect size (partial η

= .01). In other words, the

order of recalled blunt ends was not affected by whether there was blunt end blaming towards the blunt end factors present (M = -0.24, SD = 0.89) or absent (M = -0.03, SD = 0.96) in the original disaster stories. Therefore, the third hypothesis is rejected.

Blunt end recall

Regarding blunt end recall, a significant main effect of time was found [F(1.69, 133.79) = 103.26, p = .000], with a large effect size (partial η

= .57), with the recall dropping from immediate recall (M = .53, SD = .18) to one week delayed recall (M = .38, SD = .18) and stabilizing with three weeks delayed recall (M = .35, SD =.18).

The fourth hypothesis was that participants who read texts without sharp ends will recall more blunt ends than participants who read texts with sharp ends. With the factorial ANOVA test of between-subject effects, no statistically significant difference of blunt end recall in terms of number of sharp ends was found [F(1,79) = 2.77, p=.100] with a small to medium effect size (partial η

= .03). In other words, the recall of blunt ends was not affected by whether sharp end factors were present (M = .39, SD = .18) or absent (M = .45, SD = .14) in the original disaster stories that participants read in the first session, see Table 4. Therefore, the fourth hypothesis is rejected.

The fifth hypothesis was that the recall of blunt ends will be higher for participants who read texts with blunt end blaming at the end than for participants who read texts without blunt end blaming at the end. With the factorial ANOVA test of between-subjects effects, a statistically significant difference of blunt end recall in terms of blunt end blaming was found [F(1, 79) = 3.96, p=.050] with a small to medium effect size (partial η

= .05). In other words, the recall of blunt ends was affected by whether there was blunt end blaming mentioned or not in the original disaster stories. Contrary to our hypothesis, it was found that participants who read texts without blunt end blaming at the end recalled significantly more blunt ends on average (M = .46, SD = .16) than participants who read texts with blunt end blaming at the end (M = .38, SD = .16), see Table 4.

Therefore, the fifth hypothesis is rejected.

Table 4

Means and standard deviations of blunt end recall in percentages per condition

Blunt end recall

M SD

Sharp ends Present .39 .18

Absent .45 .14

Blunt end blaming Present .38 .16

Absent .46 .16

Blunt end recall over the weeks

With the sixth hypothesis, it was explored whether blunt end recall changes over time with sharp end presence or absence. With the seventh hypothesis, it was explored whether blunt end recall changes over time with blunt end blaming presence or absence. According to the repeated measures ANOVA’s with a Greenhouse-Geisser correction, no statistically significant interaction effect was found, see Table 5. Thus, the effect of number of sharp ends, as well as blunt end blaming presence or absence on blunt end recall, stays the same over the three weeks. Therefore, both hypotheses six and seven are rejected.

Table 5

Means and standard deviations of blunt end recall per manipulation from the factorial repeated measures ANOVA of between-subject effects

Blunt end recall (over time) Immediate One week

delayed

Three weeks delayed

M SD M SD M SD F p Partial η

Sharp ends

Present .50 .20 .35 .42 .33 .19 0.10 .873 .00 Absent .56 .16 .19 .15 .38 .16

Blunt end blaming

Present .49 .18 .35 .18 .31 .17 0.53 .592 .01

Absent .56 .18 .41 .17 .40 .17

Blunt end order recall

The eighth hypothesis was that the recall order of blunt ends will differ between participants who read texts with sharp ends and participants who read texts without sharp ends. The ninth hypothesis was that the recall order of blunt ends will differ between participants who read texts with blunt end blaming at the end and participants who read texts without blunt end blaming at the end. As can be seen in Table 6, neither sharp ends nor blunt end blaming had a significant effect on the blunt end recall order. Therefore, both hypotheses eight and nine are rejected.

Table 6

Means and standard deviations of blunt end recall order per manipulation from the factorial ANOVA of between-subject effects

Blunt end recall order

M SD F p Partial η

Sharp ends

Present - 1.28 1.85 0.25 .246 .02

Absent - 0.76 2.02 Blunt end

blaming

Present - 1.22 1.83 0.77 .384 .01

Absent - 0.85 2.05

Note. The mean represents the averaged delta differences, thus how many sentences earlier (positive value) or later (negative value) the blunt ends were recalled compared to the original story order.

Hypotheses blaming tendency (10-13) Assigned sharp end blame

The tenth hypothesis was that the assigned blame towards sharp end factors will be higher for participants who read texts with sharp ends than for participants who read texts without sharp ends. With the factorial ANOVA test of between-subject effect, a statistically significant difference of assigned sharp end blame in terms of number of sharp ends was found [F(1, 79) = 15.75, p

<.001] with a large effect size (partial η

= .17). In other words, the assigned sharp end blame was

significantly higher for participants who read texts with sharp ends (M = 2.70, SD = 0.66) than for

participants who read texts without sharp ends (M = 1.95 SD = 1.04), see Table 7. Therefore, the tenth hypothesis is accepted.

The eleventh hypothesis was that the assigned blame towards sharp end factors will be higher for participants who read texts without blunt end blaming at the end than for participants who read texts with blunt end blaming at the end. With the factorial ANOVA test of between- subject effect, a statistically significant difference of assigned sharp end blame in terms of blunt end blaming was found [F(1, 79) = 4.025, p = .048] with a small to medium effect size (partial η

= .05). In other words, the assigned sharp end blame was significantly higher for participants who read texts without blunt end blaming at the end (M = 2.52, SD = 0.96) than for participants who read texts with blunt end blaming (M = 2.15, SD = 0.88), see Table 7. Therefore, the eleventh hypothesis is accepted.

Table 7

Means and standard deviations of sharp end blame per condition

Sharp end blame

M SD

Sharp ends Present 2.70 0.66

Absent 1.95 1.04

Blunt end blaming Present 2.15 0.88

Absent 2.52 0.96

Note. The means can range from values 0-5. 0=not applicable, 1=not at all responsible, 2=slightly responsible, 3=moderately responsible, 4=very responsible, 5=extremely responsible.

Assigned blunt end blame

The twelfth hypothesis was that the assigned blame towards blunt end factors will be higher for participants who read texts without sharp ends than for participants who read texts with sharp ends. With the factorial ANOVA test of between-subject effect, a statistically significant difference of assigned blunt end blame in terms of number of sharp ends was found [F(1, 79) = 4.83, p = .031]

with a small to medium effect size (partial η

= .06). In other words, the assigned blunt end blame

was higher for participants who read texts without sharp ends (M = 4.01 SD = 0.61) than for